Squeezelite on MIPS (OpenWrt) stops playing & generates noise after random periods

[Braklet]

Hi Ralph, I've been trying to track down a problem for some time, this has happened with multiple versions of Squeezelite. I run SL on an Atheros AR9331 (MIPS) based device with a custom OpenWrt build:

https://wiki.openwrt.org/toh/tp-link/tl-mr3020

This runs in sync with another SL instance running on my LMS server. Occasionally, and at random periods, the MR3020 Squeezelite will stop playing and generate an irregular stutter which sounds like a note or two of the current song. I can easily recover by remotely logging in to the MR3020 and manually restarting Squeezelite. It seems to happen often after I've left the players paused overnight, but can happen in as little as 15 minutes of playing.

I set up logging and was able to capture debug for two separate failures. Around the failure time, I notice that "_output_frames" skips an inordinate number of frames.

Failure 1:

[22:32:24.308674] _output_frames:70 skip 352535 of 352535 frames
[22:32:25.007738] sendSTAT:182 STAT: STMt
[22:32:25.998458] process:512 strm
[22:32:25.998947] process_strm:267 strm command a
[22:32:25.999298] process_strm:313 skip ahead interval: 8500
[22:32:26.995036] process:512 strm
[22:32:26.995523] process_strm:267 strm command t
[22:32:26.995883] sendSTAT:182 STAT: STMt
[22:32:30.994361] process:512 strm
[22:32:30.994853] process_strm:267 strm command t
[22:32:30.995210] sendSTAT:182 STAT: STMt
[22:32:34.994357] process:512 strm
[22:32:34.994846] process_strm:267 strm command t
[22:32:34.995205] sendSTAT:182 STAT: STMt
[22:32:38.995791] process:512 strm
[22:32:38.996282] process_strm:267 strm command t
[22:32:38.996778] sendSTAT:182 STAT: STMt
[22:32:42.549536] _output_frames:70 skip 374850 of 374850 frames
[22:32:43.002954] sendSTAT:182 STAT: STMt
[22:32:43.003801] process:512 strm
[22:32:43.004513] process_strm:267 strm command t

Failure 2:

[08:57:46.181698] _output_frames:70 skip 417024 of 432797 frames
[08:57:47.006983] sendSTAT:182 STAT: STMt
[08:57:48.001203] process:512 strm
[08:57:48.001689] process_strm:267 strm command a
[08:57:48.002151] process_strm:313 skip ahead interval: 9338
[08:57:49.999632] process:512 strm
[08:57:50.000117] process_strm:267 strm command t
[08:57:50.002517] sendSTAT:182 STAT: STMt
[08:57:50.003287] process:512 strm
[08:57:50.003718] process_strm:267 strm command a
[08:57:50.004061] process_strm:313 skip ahead interval: 9339
[08:57:53.999234] process:512 strm
[08:57:53.999724] process_strm:267 strm command t
[08:57:54.000074] sendSTAT:182 STAT: STMt
[08:57:58.003774] process:512 strm
[08:57:58.004266] process_strm:267 strm command t
[08:57:58.004619] sendSTAT:182 STAT: STMt
[08:58:01.999483] process:512 strm
[08:58:01.999971] process_strm:267 strm command t
[08:58:02.000324] sendSTAT:182 STAT: STMt
[08:58:04.400775] _output_frames:70 skip 411849 of 411849 frames

At all other times, the _output_frames skip value is between 1000-3000 frames.

Any ideas about what might be happening?

My failure sounds very similar to this reported back in 2013:

http://forums.slimdevices.com/showthread.php?99916-workaround-for-pluginplayer-plays-quot-machingunsound-when-was-idle-some-time

That guy created a workaround script, restarting Squeezelite when remaining playtime gets stuck. But I'd rather figure out how to avoid the problem altogether.

Thanks for any help you can provide! (I attached full debug for both the failures, just in case.)
mr3020-squeezelite-fail.zip

BL

[Braklet]

Current running SL version is 1.8.4-743.

The MR3020 does not peg CPU nor run out of RAM, doesn't seem related to any platform resources.

[ralph-irving]

There have been a lot of changes since that version. The current version is 1.8.6-997. The issue might have been fixed although there's no specific change that I'm aware of the could have fixed it

[Braklet]

Understood. Unfortunately, OpenWrt applies a bunch of patches to SL so upgrades aren't exactly simple:

https://github.com/openwrt/packages/tree/master/sound/squeezelite/patches

I'll have to sort through all their patches, figure out what was intended, then modify them to work with Git latest.